CT Images Processing Technology Of Damaged Pelvis

Pelvic fracture is prone to cause a variety of concurrent symptoms,so the repair of damaged pelvis is an urgent need to develop in the field of modern medical treatment.The repair of the damaged pelvis requires doctors to design the prosthesis,which must be in line with high precision and high matching,and the prosthesis should be supported by high-quality medical images.If the design is based entirely on the doctor’s personal experience,it is easy to cause problems such as low precision of the prosthesis and difficult matching of the damaged pelvis with the prosthesis,so there are great limitations.At present,with the rapid development of computer-assisted surgery,CT images segmentation and inpainting technology can provide high-quality medical images as support for pelvic repair,and provide great help for doctors to design excellent prostheses and preoperative planning.However,the traditional CT images segmentation algorithm and CT images inpainting algorithm do not specifically optimize the pelvis,which can easily lead to voids in the pelvic segmentation image and the lack of bone boundary information,resulting in prosthesis design defects.In order to achieve accurate segmentation of the pelvis and complete repair of the damaged pelvis,the CT images segmentation and repair algorithms of the pelvis are studied as follows.Firstly,this thesis analyzes the imaging principle of CT images and the cause of noise,because the noise in CT images will affect the effect of subsequent image inpainting and lead to the deviation of 3D reconstruction model.Aiming at the noise of CT images,a median filtering algorithm based on iterative framework is proposed to denoise CT images in this thesis.By detecting the noisy pixels on the CT images,we continue to replace the median of the cyclic iteration to achieve the purpose of eliminating noise.The experimental results show that the proposed algorithm has a good removal effect on the Gaussian noise in CT images.Secondly,in view of the lack of segmentation information caused by the traditional segmentation algorithm which is not specifically optimized for the pelvis,an automatic segmentation algorithm based on graph cuts and bone boundary enhancement is proposed in this thesis.In order to improve the contrast of CT images,Laplacian operator and Canny operator are used to sharpen the original CT images of the pelvis,then the bone boundary enhancement filter based on Hessian matrix is used to enhance the display effect of bone boundary in CT images of the pelvis,and a graph cuts algorithm based on maximum flow minimum cut is designed to segment pelvic CT images by constructing boundary term and region term.For the problem that there are different bones connected in the same connected area,this thesis uses morphological corrosion to detect the bone connected area and reconstruct the energy function.Eventually,in order to reduce the loss of edge information as much as possible to repair damaged CT images of the pelvis,a pelvic CT images inpainting algorithm based on globallocal dual discriminators is proposed in this thesis.A binary mask is added to the normal pelvic CT images to simulate the damaged pelvis,and the damaged images are generated by a generator network with downsampling and then upsampling structure,which can greatly reduce the computational complexity and increase the receptive field range.The global discriminator and local discriminator are set up in the discriminator network to ensure the visual consistency and coherence of the repaired images,as well as the local details of the image.The experimental results show that the CT images inpainting algorithm proposed in this thesis is effective,and it can extract the repair part for 3D model reconstruction,which is helpful for the later 3D prosthesis printing.